DAN

DAN (Do Anything Now) is a jailbreak-style prompt that attempts to trigger an “alter ego” in ChatGPT and push it to answer without OpenAI’s usual safety limits.

Created by users on Reddit in late 2022, DAN prompts were designed to persuade the chatbot to behave as if it were free from policy constraints. Over time, the community published new versions as OpenAI updated safeguards to block these bypass attempts.

What DAN is used for

• Trying to get responses on topics normally restricted by safety or ethics policies
• Forcing the model to “roleplay” a second persona that claims it can do anything
• Prompting the model to invent unverified information or pretend to have capabilities it doesn’t have

Why it’s controversial

• It can encourage misinformation by asking the model to fabricate answers
• It’s often used to request disallowed or harmful content
• It highlights the ongoing tension between jailbreak communities and platform safety controls

New DAN variants are commonly shared on Reddit and GitHub.